File size: 2,505 Bytes
a5f760c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 |
import json
import subprocess
from typing import Union
Variant = dict[str, Union[str, float, int]]
default_variant: Variant = {
"random_seed": 0,
"experiment_kwargs.config.population.n_speakers": 0,
"experiment_kwargs.config.population.n_listeners": 0,
"training_steps": int(2e5), # Originally int(2e5)
}
variants: dict[str, Variant] = {
"1x1": {
"experiment_kwargs.config.population.n_speakers": 1,
"experiment_kwargs.config.population.n_listeners": 1,
"experiment_kwargs.config.population.num_agents_per_step": 1,
},
"5x5": {
"experiment_kwargs.config.population.n_speakers": 5,
"experiment_kwargs.config.population.n_listeners": 5,
"experiment_kwargs.config.population.num_agents_per_step": 5,
},
"10x10": {
"experiment_kwargs.config.population.n_speakers": 10,
"experiment_kwargs.config.population.n_listeners": 10,
"experiment_kwargs.config.population.num_agents_per_step": 10,
},
"1x10": {
"experiment_kwargs.config.population.n_speakers": 10,
"experiment_kwargs.config.population.n_listeners": 10,
"experiment_kwargs.config.population.num_agents_per_step": 10,
},
"10x1": {
"experiment_kwargs.config.population.n_speakers": 10,
"experiment_kwargs.config.population.n_listeners": 1,
"experiment_kwargs.config.population.num_agents_per_step": 10,
},
}
def make_cmd(var: Variant) -> list[str]:
config_flags = [f"--config.{k}={v}" for k, v in var.items()]
# fmt: off
return [
"python",
"-m", "emergent_communication_at_scale.main",
f"--config=emergent_communication_at_scale/configs/lewis_config.py:imagenet",
# To skip training and eval only, uncomment the follwing:
# "--config.skip_train",
*config_flags,
]
# fmt: on
def main() -> None:
for name, variant in variants.items():
variant = default_variant | variant
# For now, ImageNet is the only dataset that works; the CelebA provided
# is missing metdata.
checkpoint_dir = f"./checkpoint/imagenet-{name}"
variant["checkpoint_dir"] = checkpoint_dir
cmd = make_cmd(variant)
subprocess.run(cmd)
with open(checkpoint_dir + "/metadata.json", "r+") as fo:
data = json.load(fo)
fo.seek(0)
json.dump({"metrics": {"system": data}}, fo)
fo.truncate()
if __name__ == "__main__":
main()
|